Intel Pentium III E, S370, 900 MHz, 100MHz fsb, i810
Creative Soundblaster Live CT4620
Ati Rage Pro PCI 8mb (xpert@play, 4 mb on board + 4 mb upgrade module) driver J5.30.1-B02

40.2 fps, 31.3 seconds

Try my config. Might hit 60 - but it'll look like doom

Spark wrote on 2025-04-14, 18:38:

Intel Pentium III E, S370, 900 MHz, 100MHz fsb, i810 Creative Soundblaster Live CT4620 Ati Rage Pro PCI 8mb (xpert@play, 4 mb on […]
Show full quote

Intel Pentium III E, S370, 900 MHz, 100MHz fsb, i810
Creative Soundblaster Live CT4620
Ati Rage Pro PCI 8mb (xpert@play, 4 mb on board + 4 mb upgrade module) driver J5.30.1-B02

40.2 fps, 31.3 seconds

You can try my newer 4.11.2611 driver and use included .bat file before or after driver install. Have not hand picked OpenGL or crazy edited atibdg.ini file. Newer driver and core D3D components are as fast as possible and from that i already get my new 3dmark record soon or better OpenGL results if i want. Not any big change, but still, bit by bit.Also Rage3 ATiDVD playback now supported in the 3DCIF driver package.

Hmm... now I'm curious. I have a rather rare 3D accelerator from 1995 that apparently has OpenGL support in Windows NT. I wonder what kind of numbers it'd pull.

It is an AccelGraphics AG300, like the one on this page. It uses an LSI L1B9307 chip (absolutely huge thing with a ridiculous number of pins) and has 5MB VRAM for a frame buffer and 2.5MB for a Z-buffer. No texture mapping support, so I'm sure the speed will be horrendous and probably rely a lot on the CPU.

EDIT: I have no idea why I felt the need to do this, but I believe that crazy huge LSI chip is a 304-pin QFP. I counted 76 pins per side. @_@

Does this Quake3 demo needs something else, could not load default.cfg error. More used Q1 and Q2 for testing.
I have 3D Rage Pro AGP 2x with slow 4MB EDO, still should be faster than any Rage2 or S3 Virge cards out there.

Exactly this oldie, board reads 1997, can give this a try if Q3 demo will start, let see.

Q3 demo is 1.11 and won't involve the FOUR demo this test calls for (especially as there's been very regressive protocol updates between 1.11 and 1.25 and the test involves a map not in the demo). You must have a standard full version of Q3A.

Duron 650Mhz/128MB/W98SE DX6.1/cmi8738/viadma missing@now
3D Rage Pro 2x AGP 4MB EDO 54.1fps k6q3 patch / without patch 53.2 (1997 at board) 538x3D2611 driver @ OpenGL 1091
ATi Rage XL 2xAGP 8MB RAM 61.1fps k6q3 patch / without patch 57.7 (1999 at board) 538x3D2611 driver @ OpenGL 1091

Intel Pentium III E, S370, 900 MHz, 100MHz fsb, i810
Creative Soundblaster Live CT4620
i810 AGP 0mb (12mb agp mem)

53.8 fps, 23.4 seconds

Incredibly, this only just beats the OP's score for a 4mb PCI Riva 128 (51.3 fps)
No wonder the bad reputation.

Spark wrote on 2025-04-15, 20:15:

Intel Pentium III E, S370, 900 MHz, 100MHz fsb, i810 Creative Soundblaster Live CT4620 i810 AGP 0mb (12mb agp mem) […]
Show full quote

Intel Pentium III E, S370, 900 MHz, 100MHz fsb, i810
Creative Soundblaster Live CT4620
i810 AGP 0mb (12mb agp mem)

53.8 fps, 23.4 seconds

Incredibly, this only just beats the OP's score for a 4mb PCI Riva 128 (51.3 fps)
No wonder the bad reputation.

Amazing 0.3fps more with rage pro 4mb edo agp, with rage xl 8mb agp advantage already visible 53.8fps vs 61.1fps.
Here we should get lower the better , not higher the better. Changed score atitude, lower is the new best of the best.

Thanks for benchmarking, guys! 😀
I'm still waiting for the delivery of a Rage II (non +) with 4MB SGRAM. I'll then add your results to the table as well.

There's no way to display tables here in the forum, is there?
Currently I have to print Excel tables as images. That is quite cumbersome.

Print as image?
Good old snipping tool no good?

RaVeNsClaw wrote on 2025-04-16, 08:50:

Thanks for benchmarking, guys! :-) I'm still waiting for the delivery of a Rage II (non +) with 4MB SGRAM. I'll then add your re […]
Show full quote

Thanks for benchmarking, guys! 😀
I'm still waiting for the delivery of a Rage II (non +) with 4MB SGRAM. I'll then add your results to the table as well.

There's no way to display tables here in the forum, is there?
Currently I have to print Excel tables as images. That is quite cumbersome.

Defining text as a code block will use a font with fixed spacing. ASCII tables will then be aligned. To do the conversion, you could try a tool such as https://tableconvert.com/excel-to-ascii (just an example).

RaVeNsClaw wrote on 2025-04-16, 08:50:

Thanks for benchmarking, guys! :-) I'm still waiting for the delivery of a Rage II (non +) with 4MB SGRAM. I'll then add your re […]
Show full quote

Thanks for benchmarking, guys! 😀
I'm still waiting for the delivery of a Rage II (non +) with 4MB SGRAM. I'll then add your results to the table as well.

There's no way to display tables here in the forum, is there?
Currently I have to print Excel tables as images. That is quite cumbersome.

For Rage Pro/XL cards works well GLDirect 3.0, even OpenGL 8-bit paletted texures working and only for that i am using it (D3D 8bit must work).
Today or tomorrow i try Rage2C AGP 4MB EDO vs Rage2C AGP 4MB SDR. I also have SIS305 PCI ,Savage4 8MB, S3 Virge GX2, S Trio3D/2X +more.
Rage2C is Rage2 based AGP, slow as other Rage2 cards, Rage2 that i like bit more is ATi Rage2+DVD, i would use it more as 2D card, than 3D card.

The Rage II arrived and I've added the entries!
The performance is almost identical to Rage II+, but it has more optical glitches. 😉

Ozzuneoj wrote on 2025-04-15, 02:29:

Hmm... now I'm curious. I have a rather rare 3D accelerator from 1995 that apparently has OpenGL support in Windows NT. I wonder what kind of numbers it'd pull.

It is an AccelGraphics AG300, like the one on this page. It uses an LSI L1B9307 chip (absolutely huge thing with a ridiculous number of pins) and has 5MB VRAM for a frame buffer and 2.5MB for a Z-buffer. No texture mapping support, so I'm sure the speed will be horrendous and probably rely a lot on the CPU.

That sounds very interesting! Please test that card! 😀

Spark wrote on 2025-04-15, 20:15:

Intel Pentium III E, S370, 900 MHz, 100MHz fsb, i810 Creative Soundblaster Live CT4620 i810 AGP 0mb (12mb agp mem) […]
Show full quote

Intel Pentium III E, S370, 900 MHz, 100MHz fsb, i810
Creative Soundblaster Live CT4620
i810 AGP 0mb (12mb agp mem)

53.8 fps, 23.4 seconds

Incredibly, this only just beats the OP's score for a 4mb PCI Riva 128 (51.3 fps)
No wonder the bad reputation.

Hey, that iGPU is still faster than a Permedia 3, even without dedicated memory. 😉

marxveix wrote on 2025-04-15, 16:28:

Duron 650Mhz/128MB/W98SE DX6.1/cmi8738/viadma missing@now
3D Rage Pro 2x AGP 4MB EDO 54.1fps k6q3 patch / without patch 53.2 (1997 at board) 538x3D2611 driver @ OpenGL 1091
ATi Rage XL 2xAGP 8MB RAM 61.1fps k6q3 patch / without patch 57.7 (1999 at board) 538x3D2611 driver @ OpenGL 1091

Sorry, I only understood half of that.
What is the k6q3 patch?

marxveix wrote on 2025-04-15, 16:28:

Duron 650Mhz/128MB/W98SE DX6.1/cmi8738/viadma missing@now
3D Rage Pro 2x AGP 4MB EDO 54.1fps k6q3 patch / without patch 53.2 (1997 at board) 538x3D2611 driver @ OpenGL 1091
ATi Rage XL 2xAGP 8MB RAM 61.1fps k6q3 patch / without patch 57.7 (1999 at board) 538x3D2611 driver @ OpenGL 1091

RaVeNsClaw wrote:

Sorry, I only understood half of that.
What is the k6q3 patch?

Its AMD K6 3DNow for Q3, Q3 may say its using 3DNow without it, but it aint or doing it very poorly. Q3 uses SSE better. You dont need K6-2 or K6-3, just CPU with 3Dnow. Your CPU supports SSE, use that, my Duron has 3DNow and no SSE. Similar case with Quake2 you needed later version of Quake2 or AMD update. Even Quake1 had one build with 3Dnow support, but nobody has it, it called ppquake. Quake3 and 3DNow cpu owner copy extracted files from q3k6-132opt.exe to baseq3 and you will like it, oh yes dllset.bat click@done.

Thanks! Didn't know that there is a 3DNow patch!
Man, it's been so long that I heard of 3DNow.

It's much less a '3dnow patch' and more of a bunch of non-QVM game modules. QVM's only a slight CPU hit compared to native code, and Q3 dropped having native code modules a long time ago for the sake of portability. There's no vital changes in rendering with this "3dnow patch".

leileilol wrote on 2025-04-17, 00:09:

It's much less a '3dnow patch' and more of a bunch of non-QVM game modules. QVM's only a slight CPU hit compared to native code, and Q3 dropped having native code modules a long time ago for the sake of portability. There's no vital changes in rendering with this "3dnow patch".

Thank you, performance difference is there and its made for 3DNow. 😀

The page where it came from doesn't even mention 3dNow! anywhere. Nothing is disclosed on how the games got optimized; for all you know these could be just gcc -mtune -flto compiles with a prayer.

The same page briefly had SOF2 DLLs but were removed because they didn't make any difference. Know why? SOF2 wasn't using QVM in the first place, so there's no '3dnow performance boost' to be had. The game/cgame module can't touch the renderer.

leileilol wrote on 2025-04-17, 06:28:

The page where it came from doesn't even mention 3dNow! anywhere. Nothing is disclosed on how the games got optimized; for all you know these could be just gcc -mtune -flto compiles with a prayer.

The same page briefly had SOF2 DLLs but were removed because they didn't make any difference. Know why? SOF2 wasn't using QVM in the first place, so there's no '3dnow performance boost' to be had. The game/cgame module can't touch the renderer.

Thank you for explaining, xp version dint work with Duron, i had to try.
q3k6-132opt was 0.3-0.4fps faster than q3tb-132opt with my Duron.
Duron 650Mhz/128MB/W98SE DX6.1/cmi8738/viadma missing@now
ATi Rage XL 2xAGP 8MB RAM 61.2fps q3k6-132opt / (1999 at board) 538x3D2611 driver @ OpenGL 1091
ATi Rage XL 2xAGP 8MB RAM 60.9fps q3tb-132opt / (1999 at board) 538x3D2611 driver @ OpenGL 1091
Run tests 3 times to be sure, R3-XL 1260 frames, 20.6 seonds, 61.2fps